Usability of semi-formal and formal methods integration : empirical assessments

FACULTY OF ENGINEERING, SCIENCE AND MATHEMATICS SCHOOL OF ELECTRONICS AND COMPUTER SCIENCE Doctor of Philosophy by Rozilawati Razali Software practitioners are provided with an enormous choice of methods and tools for improving software quality. They seem to adopt a new technology only if there is convincing evidence that the technology is usable. Furthermore, it is no longer acceptable in software engineering to claim that a new technology is usable without accompanying scientific evidence. Empirical assessments provide an ideal mechanism for evaluating software technology. As a single assessment can never embrace all possible situations, it is reasonable to acquire some evidence of a new technology’s likely efficacy when used under certain conditions. The use of formal notations such as B increases a model’s precision and consistency. However, the notations are regarded as being difficult to comprehend due to unfamiliar symbols and underlying rules of interpretation that are not so apparent to practitioners. Semi-formal notations such as Unified Modelling Language (UML) use graphical representations to present system elements. They therefore are perceived as more accessible. Nevertheless, such notations cannot be verified systematically to ensure the correctness of a model. Perhaps by combining both semi-formal and formal notations could produce a model that is not only accurate and consistent but also accessible to practitioners. This thesis presents several empirical assessments conducted on a modelling method that integrates the use of B formal notation and the semi-formal notation of UML, namely UML-B. The main objective of the assessments is to evaluate the usability of the method. This includes the comprehensibility, learnability, operability and attractiveness of the notation used in the method and the method itself in supporting modelling process. The assessments comprise a series of controlled experiments and surveys. The controlled experiments evaluate the comprehensibility of the notation from stakeholders’ perspective for model validation and maintenance tasks. On the other hand, the surveys assess the usability of the method and the supporting tools from developers’ perspective for model creation tasks. The findings of the assessments suggest that the method is able to produce a comprehensible formal model. The method is accessible to users only when the principles and roles of both notations are obvious and well understood, and when there is strong support from the environment. Based on the findings, a set of usability theories of integrated methods such as UML-B has been generated. Design profiles have also been proposed, which provide future designers with some guidelines for designing usable integrated methods and verification tools. The thesis also provides practitioners with some understanding of the strengths, weaknesses, opportunities and threats of methods that integrate semi-formal and formal approaches.